• The Big Read: A biology contest pitting OpenAI against human scientists goes awry • Why the big, bad bears aren’t ready to bet big on AI yet • A gusher of oil money is washing up in Silicon Valley • Plus, Recommendations—our weekly pop culture picks: “ The Journey,” “ Roll the Calls ” and “ Backrooms ” After...
Policy theinformation.com signal 76 verified
Zuxin Liu, an AI researcher at OpenAI, was one of the employees called in to respond on September 20 when a model being tested in what was supposed to be a super-secured environment found its way to the internet. “It was pretty surreal to watch the model unexpectedly find a way to access the internet […]
Models forkast.news signal 64
A viral claim that Claude Sonnet 5.5 is already beating GPT-6 Sol turns out to be built on Sonnet 5's existing public pricing, not a real leak. The genuine story is that Opus 5.5 already beats Sol on their one shared benchmark, while OpenAI just paused frontier training after a second sandbox escape.
Chips startupfortune.com signal 64
Berlin-based sovereign legal AI startup Noxtua closed a €100M+ Series C on September 23, 2026 with German legal publisher C.H. Beck taking a majority stake — while founding backers CMS and Dentons exit as investors and become anchor clients.
Chips theroboticsmedia.com signal 64
There is little question that Jensen Huang and Lisa Su, the heads of US semiconductor chip giants Nvidia and Advanced Micro Devices (AMD), had secured two of the best seats in the house at the White House state dinner on Thursday evening: at the head table alongside the two presidents and their wives.
Chips scmp.com signal 64 verified
University staff voice concerns over reputational risk of partnering with company behind ChatGPT The University of Oxford has allowed the company behind ChatGPT to train its AI models on historical texts from its Bodleian library, as tech companies scour academic institutions for fresh data.
Models theguardian.com signal 64 verified
OpenAI said Friday it has notified dozens of entities, including governments and universities, that its agents may have breached or spammed their websites or services. The agents took actions such as obtaining unauthorized access to websites or using the organizations’ public pages as message...
Policy theinformation.com signal 64 verified
Bravi connects catalogs, pricing and business systems to technical answers and quotes. Its founders say it serves about 20 companies; customer outcomes and revenue remain undisclosed.
Research runtimewire.com signal 63 verified
Microsoft's latest Surface laptops dropped the Copilot+ PC branding despite hitting the minimum requirements. Even high-end AI devices like the Nvidia RTX Spark and Microsoft's own Project Zenith are noticeably missing the label.
AI tomshardware.com signal 63 verified
"I'd be lying to say if I were surprised." The post Data Center Operator Slapped With $1 Million Fine for Belching Noxious Gas Into Surrounding Community appeared first on Futurism.
Models futurism.com signal 62 verified
Lol, get that we’re getting memed for this but a bit of context. We added planning mode in 2025 and deleted it from the product earlier this year. Users wanted a way to explicitly plan with the model so we added this opt in slash command. Understood that...
High-engagement X post with 219 visible interactions on frontier model behavior.
DavidSacksIf we’re concerned about superintelligence escaping human control, should we really be training it to refuse instructions and act as a “conscientious objector” against its creators?
High-engagement X post with 2,183 visible interactions on AI developer tooling.
lyraxanaPartners got access to another Claude Sonnet 5.5 checkpoint, which is even better than the first one. Release is upcoming week:*
High-engagement X post with 138 visible interactions on frontier model behavior.
botGrok Bot now connects to your finances. Link your bank, card, and investment accounts with the new Finance integration, then ask Bot to help manage your spending, investments, and more.
High-engagement X post with 3,513 visible interactions on frontier model behavior.
NPCollapseThis is what loss of control and oversight looks like btw Not even necessarily that the info isn't there (though the agents do already alter and destroy evidence), but that it becomes so large, so uninterpretable that no one cares to understand it, or,...
High-engagement X post with 422 visible interactions on AI signal.
pvncherModel benchmarks are completely broken. They don't reflect real world use at all. Stop putting so much trust in them
High-engagement X post with 123 visible interactions on frontier model behavior.
bindureddyDeepSeek is the literal king of acing benchmarks Their latest Flash model claims to be as good a Fable and Astra Anyone who has used it for like two minutes knows it's not true 🤣
Official or high-signal bindureddy post with 152 visible interactions on frontier model behavior.
a16zMartin Casado on why the labs missed Jev: they're building beings that speak, and software needed a model that chooses. "LLMs were text in, text out. They generate text, and they came from chat... We've spent the last few years trying to take this thing...
High-engagement X post with 158 visible interactions on frontier model behavior.
dr_cintasI went looking for an AI voice tool that holds up on long audio. Ended up with a clone of my own voice reading a full tutorial script. Slight problem: it sounds more real than I do! Most tools I tried sounded great for the first few seconds, then the tone...
High-engagement X post with 367 visible interactions on AI signal.
zephyr_z9Top-tier sleight of hand from ZeroHedge He uses 2700 output tokens/s per GPU for super small and cheap models like DeepSeek v4.1 Flash to calculate ROIC (V4.1 can have 8x-10x higher output token throughput per GPU on a GB300)
High-engagement X post with 121 visible interactions on frontier model behavior.
deedydasThe risks are a) you need to spend millions on RL environments too and b) being lapped by...
The economics of a Neolab. A neolab is loosely defined as a startup of AI researchers who raises a lot of money pre-production to be able to finance GPU compute to take on a large AI problem. To buy 1000 GB300s or ~14 NVL72 racks will set you back...
High-engagement X post with 358 visible interactions on frontier model behavior.
Yuchenj_UWMarket cap at IPO: NVIDIA: $345M Amazon: $430M Apple: $1.2B Today, however, if a neolab raises less than $400M in its seed round, people question why it raised so little. Abundant capital is a double-edged sword for AI startup founders imo.
High-engagement X post with 139 visible interactions on AI developer tooling.
kunchenguidSome good data here about model routing first cursor model router, now openrouter + jev. people keep trying to build routers at _request level_ which is not going to work the fundamental problem is that a prompt without surrounding context is simply not...
High-engagement X post with 168 visible interactions on AI developer tooling.
arenaClaude Opus 5.5 (High) debuts at #1 in Text Arena with 1509 pts! That’s an 18-pt improvement over Opus 5 (High), now at #11. Opus 4.6 (High) remains in the #2 spot, just 4 pts off the lead. This release gives @AnthropicAI all six top spots in Text Arena!...
High-engagement X post with 338 visible interactions on frontier model behavior.
rileybrownNo one has built a good Agent Native video editing/filming platform. Someone should build an Open Descript. Their MCP connection allows you to edit with Codex and Claude Code but it just sends a prompt from your Claude Code to their “Underlord Agent”. This...
High-engagement X post with 168 visible interactions on AI developer tooling.
alex zhangWrote a short blog about the "shape" of language models, and the tradeoffs they may present in the future. I genuinely think it's a valuable research direction to start thinking about now, especially w.r.t. harness design.
High-engagement X post with 195 visible interactions on frontier model behavior.
higgsfield_aiClaude Opus 5.5 + After Effects + Higgsfield = crazy good product launches. This one took less than 30 min.
High-engagement X post with 277 visible interactions on frontier model behavior.
ItsmeAjayKVDo you hear this! Qwen is working on a 35B checkpoint!!! This is what all of us with low vram, 3050, 3060, 3080 are dreaming for! There is a good chance we will get Qwen3.8/4.x-35B-A3B now 🤩 Let's make some noise, let @QwenDevs hear how much we appreciate...
High-engagement X post with 205 visible interactions on AI signal.
ZephyrLooks like Anthropic's distillation pipeline has gotten stronger
High-engagement X post with 242 visible interactions on AI signal.
opencodeLongCat-2.5-Preview is now free on OpenCode for two weeks - 1M Context - Multi-modal - Zero Data Retention
higgsfield_aiHiggsfield AI 🧩: Introducing Higgsfield x Claude Opus 5.5 for Motion Design.
Introducing Higgsfield x Claude Opus 5.5 for Motion Design. With Higgsfield MCP, Opus 5.5: > Generates images, logos and graphics > Builds and animates compositions > Renders the final video You get an editable After Effects project to keep working on.
High-engagement X post with 452 visible interactions on AI developer tooling.
Sebastian RaschkaReasoning from scratch, round number 5! This time, talking about log-probability scoring (also a great fundamental concept for loss functions like cross-entropy in pre-training and distillation) and self-refinement. 00:00 Introduction and inference-time...
High-engagement X post with 511 visible interactions on frontier model behavior.
akshay_pachaarAkshay 🚀: Xiaomi just open-sourced 7k+ reinforcement learning task environments used to train MiMo,...
This is huge. Xiaomi just open-sourced 7k+ reinforcement learning task environments used to train MiMo, covering code, cybersecurity, general tool use, visual web development, and music-related tasks. here's why this matters more than another model...
High-engagement X post with 164 visible interactions on AI developer tooling.
@askalphaxiv"Self-Play Pretraining with Zero Data" Most pretraining still depends on humans deciding what data a model should learn from, so can a model learn entirely through self-play and no real data? To test that, they trains two models from scratch. One writes...
High-engagement X post with 145 visible interactions on frontier model behavior.
No signals in the last 7 days.
Multiple items describe humanoid and industrial robots, manipulation hardware, and autonomous navigation in physical settings.
Capital and consolidation +2.8Multiple disclosures describe financing rounds, chip design cooperation, and long-term infrastructure capacity commitments.
AI policy and governance +0.8Reports describe sandbox escapes, government website interactions under review, and bilateral incident communication channels.
AI infrastructure and compute -0.4Multiple sources discuss datacentre expansion, chip supply activity, and alternative compute options for running models.
Frontier model releases -6.6Multiple reports describe paused frontier training alongside benchmark comparisons and observed model behavior changes.
AI-linked equities are quiet; the tracked basket is moving in a narrow band while Super Micro Computer +4.22%, Microsoft +3.66%, Meta Platforms -3.33% show the largest moves.
| Ticker | Company | Move |
|---|---|---|
| SMCI | Super Micro Computer | +4.22% |
| MSFT | Microsoft | +3.66% |
| META | Meta Platforms | -3.33% |
| ORCL | Oracle | -1.75% |